Businesses now think beyond database administration and about intelligent decision support, online analytical processing, data warehousing, data mining, metadata, and universal data access. Or at least they should.
The elements of data alignment appear in the chart below, designed to help define your data, information, and knowledge strategies, as well as prioritize requirements.
DATA INFO KNOWLEDGE STORAGE DBMSs Data Knowledge Repositories Oracle Warehouses & Content Managers DB2/UDB SQL Server Data Marts ___________________________________________________________________ ANALYSIS On-line Data Mining Knowledge mining, transaction OLAP sharing, & processing ROLAP dissemination MOLAP Standard query ____________________________________________________________________ PRODUCTS & SERVICES
Not long ago the list of major data base vendors included five or six players, but now for all practical purposes only three remain (Oracle, DB2/UDB, SQL). In addition to core database management system (DBMS) applications are the hardware storage systems you'll need to process your business data, information, and knowledge. Another major trend is the movement from hierarchical to relational database management systems and the migration from relational to object-oriented database management. The more distributed your applications -- and the greater your need for flexibility -- the more you'll want to move to an object data architecture. Avoid supporting even two database environments; if its at all possible select one.
Information storage options -- data warehouses, data marts and special purpose hybrids -- require serious thinking about where your business is headed, how much money is available for the construction of these artifacts, what users will require, and (this is related) what your data mining tools will look like.
So what should you spend money on?
If you're a DB2/UDB or Oracle shop and your business model calls for data warehousing with front-end analysis tools, you may need to implement an entire blueprint. But if your needs are more modest, you can pick and choose. A note of caution: everything needs to work together. If you don't have cracker-jack data architects on your staff, go out and get some. If you fail to implement your strategy holistically, you'll spend orders of magnitude more money (and time) than you have to.
If you've implemented an enterprise resource planning application (like SAP's R/3 or Oracle's Financials), then you'll have additional integration and interoperability requirements to satisfy.
Knowledge storage is akin to dressing for a party to which you have no directions. In a sense, you're investing in a solution in search of a problem. The serious (read: measurable) pain relieved by the "knowledge management" (KM) business is generally better described by consultant doctors than by the patients (that is, the businesses themselves). To play this game, you'll have to figure out the specific problems this amorphously defined technology can solve, then consider how to store your business' unconventional, unstructured "knowledge."
Here's some food for thought. Rather than be flip about this young field, let's look at some of its assumptions. First, KM assumes there is knowledge to manage -- that you've somehow codified the collective wisdom of your industry's and company's experiences. Second, it assumes that your culture and processes give priority to info sharing and are capable of exploiting codified knowledge. Third, it assumes that you have (or are willing to invest in) the tools to make this happen.
Some vertical industries will be in better positions to exploit knowledge management. But others will have little or no need for what the consultants are assuring us is the next great revolution in database management. Look at your industry, your culture, your processes, and your current and planned data infrastructure. If a KM system seems appropriate, run a pilot project to validate your expectations.
On a separate front, any company processing a lot of content has abundant opportunities to exploit content management tools and applications. At some point, such businesses will need to move to a serious content management platform from vendors like Interwoven and Vignette. But make sure your requirements justify the investment to acquire and support these platforms.
Storage is essential to analysis. But what the heck is this OLTP and OLAP stuff? Simply, online transaction processing (OLTP) is what everyone's been doing for a long, long time. Online analytical processing (OLAP) -- especially when coupled with data warehousing technology -- is how data, information, and knowledge get usefully exploited.
OLTP is the mother of all analysis. It provides insight into a business' internal Internet data, especially as it applies to operations and real-time transactions. OLAP can be used to provide strategic insight, to develop reports and analyses, or to make sense of data when access to it is unstructured.
It's easy to see why OLAP has fans: it provides flexible querying of data that OLTP leaves more or less untapped. In fact, OLAP provides a gateway to "information analysis."
Information analysis requirements extend OLAP's capabilities to desktop OLAP (DOLAP), relational OLAP (ROLAP) and multidimensional OLAP (MOLAP). DOLAP include PC-based tools that support the analysis of data marts and warehouses; ROLAP include server applications that support analyses based on a relational database management system or a data warehouse; and MOLAP exploits pre-developed data cubes.
Knowledge analysis and management is the end game of database management, data warehousing, OLAP, and data mining. It's also at the heart of any learning organization. But it suffers from an identify crisis and should be pursued only when the criteria described in the Knowledge Storage section above are satisfied.
Don't Forget the Plumbing...
Data, information, and knowledge infrastructure issues are complex because they are operational: all of this stuff has to work nearly all the time, which means you have to define and apply processes that can be implemented and maintained over time.
Infrastructure support can kill you as your businesses begins to connect more and more of its operations. You should perform a cost-benefit analysis of your options here. Supporting the data, information, and knowledge infrastructure in-house may give you the right control, but it also may give you headaches you'd prefer not to have.
Take a hard look at your IT employees; if you've got the right personnel, then consider the costs and benefits of in-house support. But if not, then look seriously at getting outside help.
The watchwords here are reliability, scalability, security, integration and inter-operability. If you can handle all that, terrific; if not, get some help.
Note that I never recommend outsourcing infrastructure design or architecture. Every business must develop its own requirements. Implementation and support, of course, are a different matter.
The Holy Grail
So where's all this technology leading? The major database players are scrambling to provide universal data access from all tethered and un-tethered devices. Eventually, structured, unstructured, hierarchical, relational, object-oriented data, information, and knowledge will be ubiquitously accessible.
While we're a few years away from all this, it's helpful to understand the holy grail and to adapt your business models in the general direction of universal data access. Microsoft, IBM, and Oracle all have plans to provide such access. It's important to stay abreast of their progress -- and the implications to your business models and processes.
Next month we'll look at security.
Steve Andriole is the founder & CTO of TechVestCo, a consortium that focuses on optimizing IT investments. He is a former executive of both Safeguard Scientifics and CIGNA Corporation. His career began at the Defense Advanced Research Projects Agency. He can be reached at email@example.com.